GingerBangla: A Supervised Machine Learning Approach for Predicting Ginger Yield using Bangladeshi Climate Data

Abstract

Ginger (Zingiber officinale) farming is an important part of Bangladesh's agriculture, and it also contributes significantly to the country's economy. Accurate prediction of ginger yield is essential for farmers, and decision-makers, to optimize their cultivation practices, and for other agricultural stakeholders to use in their strategic planning. In this study, We propose to use Bangladesh climate data to develop a supervised machine-learning method for predicting ginger yield. To predict ginger yield, four machine-learning models were used: XBGR, GBR, RFR, and DTR. The model's accuracy was evaluated using historical data from 23 domestic Bangladeshi regions covering a 54-year period from 1968 to 2021, obtained from the Government organization in Bangladesh. According to the results, XBGR predicted ginger yield with an accuracy of 91%, GBR, and RFR predicted it with an accuracy of 95%, and DTR predicted it with an accuracy of 85%. The results demonstrate that supervised machine learning models can predict ginger yield with accuracy using climate data. The high accuracy of the GBR and RFR models shows that they can be used to predict ginger yield in Bangladesh. These models can help farmers and policymakers make smart decisions about handling crops, where to put resources, and how to reduce risks. This study helps to address current issues with ginger farming in Bangladesh. Production of ginger is severely hampered by unpredictable climate patterns, which include temperature changes, rainfall unpredictability, and severe weather events. Using supervised machine learning models, especially GBR and RFR, to predict ginger production in Bangladesh is a good way to do this. The accurate yield predictions made possible by this study can help farmers, reduce expenses, and make sure that ginger production is sustainable, even when the weather is unpredictable.